© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Jacobian LensNEW
    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsinterp-engineNEWAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Joseph Bloom · Feature Splitting for GPT2-Small
    5. GPT2-Small
    7. Residual Feature Splitting 24576
    9. 8-RES_FS24576-JB
    11. 14021
    Prev
    Next
    INDEX
    Explanations

    prepositions or adverbs denoting a specific location or time

    oai_token-act-pair · gpt-3.5-turboTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GPT2-SMALL @ 8-res_fs24576-jb
    Configuration
    jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    Skylion007/openwebtext
    Features
    24,576
    Data Type
    torch.float32
    Hook Point
    blocks.8.hook_resid_pre
    Architecture
    standard
    Context Size
    128
    Dataset
    Skylion007/openwebtext
    Hook Point Layer
    8
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Neuron Alignment
    Index
    Value
    % of L₁
    99
    +0.15
    0.7%
    59
    +0.11
    0.5%
    293
    +0.10
    0.4%
    Correlated Neurons
    Index
    P. Corr.
    Cos Sim.
    59
    +0.15
    0.02
    99
    +0.11
    0.02
    471
    +0.10
    0.01
    Negative Logits
    alian
    -0.77
    soDeliveryDate
    -0.74
    papers
    -0.74
    onga
    -0.74
    Kids
    -0.73
    ahime
    -0.71
    utenberg
    -0.71
    sect
    -0.68
    rams
    -0.68
    SON
    -0.67
    POSITIVE LOGITS
     arrival
    1.24
     completion
    1.01
     encountering
    0.93
     receipt
    0.84
     realizing
    0.84
     entering
    0.83
     inspection
    0.81
     arriving
    0.81
     graduating
    0.81
     returning
    0.79
    Activations Density 0.033%

    No Known Activations